The cluster configuration operation is originally to make sure
the instance won't be added back automatically between removal and
check for the number of existing peer-mode instances. But this
could make some node removed before the removal command.
Use longer sync interval instead to avoid this problem.